AKASA Demonstrates Generative AI for Revenue Cycle Management

AKASA is using institution-tuned large language models for revenue-cycle management in healthcare, where patient records average 60 documents and 50,000 words, according to CEO Malinka Walaliyadde. Some clients have all hospital billing reviewed by the AI model while humans continue to review results, highlighting the importance of workflow control and human oversight in production deployments.

AKASA Demonstrates Generative AI for Revenue Cycle Management Healthcare IT Today reported on July 8, 2026 that AKASA is using institution-tuned LLMs for revenue-cycle workflows where patient records can average 60 documents and 50,000 words , according to CEO Malinka Walaliyadde. The article says some clients have all hospital billing reviewed by the AI model, while humans continue to review results. AKASA's own materials support the broader positioning: it sells generative-AI tools for coding, CDI, prior authorization, and claims workflows. For practitioners, the story is less about a new model than about production design: site-specific tuning, measurable coding accuracy, human review thresholds, and auditable billing decisions matter more than generic GenAI claims. The implementation lesson is that healthcare revenue-cycle AI depends on workflow control more than broad model branding. The same LLM that looks impressive in a demo can create billing and compliance risk unless site-specific data, evaluation, human review, and audit trails are built into the operating model. What happened Healthcare IT Today published an interview-focused article with AKASA CEO and cofounder Malinka Walaliyadde about generative AI in revenue-cycle management. The article reports that Walaliyadde said patient records average 60 documents and 50,000 words, that AKASA tunes its LLM for each institution, and that some clients have all hospital billing reviewed by the AI model while humans continue reviewing results. Technical context Revenue-cycle work involves structured codes, payer rules, clinical documentation, and site-specific billing practices. That makes a generic chatbot framing too loose. The production pattern is closer to a specialized decision-support system: models extract and reason over clinical and financial context, while deterministic checks, review queues, and human validation help control hallucination and compliance exposure. For practitioners The due-diligence questions are concrete: what training or retrieval data is institution-specific, how coding accuracy is measured, how false positives and missed charges are handled, and how decisions are logged for audit. Human review thresholds should be explicit, especially when AI touches billing workflows that affect reimbursement and patient financial records. What to watch Look for independently reported metrics on coding accuracy, recovered revenue, denial reduction, staff time saved, and error rates. Vendor claims are useful context, but healthcare buyers should prioritize validated outcomes and compliance evidence over broad generative-AI positioning. Key Points - 1AKASA frames revenue-cycle AI around institution-specific tuning, human review, and long clinical records rather than generic chatbot use. - 2Practitioners should verify coding accuracy, false-positive handling, audit logs, and human-review thresholds before production adoption. - 3Healthcare buyers need validated operational outcomes because billing AI affects reimbursement, compliance, and patient financial records. Scoring Rationale This is a practitioner-relevant healthcare AI deployment story, especially for revenue-cycle teams evaluating LLM-assisted coding and billing review. It is operationally useful but still vendor/interview-driven, so the score remains moderate. Sources Public references used for this report. Practice with real Ad Tech data 90 SQL & Python problems · 15 industry datasets Active Search Campaigns by BudgetEasy /problems/sql/active-search-campaigns-by-budget High CPC Clicks & Poor Landing PagesMedium /problems/sql/high-cpc-clicks-poor-landing-page Campaign ROAS by Attribution ModelHard /problems/sql/campaign-roas-by-attribution-model 250 free problems · No credit card See all Ad Tech problems /problems/datasets/adtech